Design Creativity 2010 part 13 pdf

Conceptual Design and Cognitive Elements of Creativity: Toward Personalized Learning Supports for Design Creativity 109

5 Results and Discussion
Forty four senior or first-year graduate students from
the Interdisciplinary Design course at the
Sungkyunkwan University participated in the
experiment. Figure 3 shows examples of pre-test (b)
and post-test (c) performed by a student. This example
shows one sample case of the enhanced design
creativity: The average score over the 5 cognitive
elements of the post-test was increased by 1.05 from
that of the pre-test. The assigned score range is
between 1 and 5 (inclusive).
Four domain experts evaluated the conceptual
design results. The Cohen’s Kappa value was
computed from the assigned scores for inter-rater
reliability. The overall Kappa value was 0.44 over the
five cognitive elements and the significance of the
acquired Kappa value is “moderate agreement.” The
individual Kappa values were 0.35, 0.66, 0.34, 0.47,
and 0.39, respectively for flexibility, fluency,
originality, elaboration, and problem sensitivity
respectively. Fluency is considered as strongly
reliable, compared to other cognitive elements.
Table 3. Paired t-test result with p-values between pre-test
and post-test data
t p-value
Fluency -4.103 0.000
Flexibility -3.197 0.003
Originality -5.367 0.000
Elaboration -0.604 0.549

Problem Sensitivity -0.623 0.537

5.1 Enhanced Design Creativity
As a result, 31 students out of 44 students showed the
enhanced design creativity with regard to the 5
cognitive elements (70% increases), possibly
indicating the effectiveness of the creativity exercise


a. Conceptual design task



b. A sample of pre-test c. A sample of post-test
Fig. 3. Cenceptual design task , and two samples of pre-test and post-test acquired from a student. Conceptual design task is
used for pre-test and post-test. Note that both samples of pre-test and post-test were evaluated by human experts
110 Y.S. Kim, J.H. Shin and Y.K. Shin

program. The overall difference between pre-test and
post-test are +0.86, +0.32, +0.65, +0.06 and +0.06,
respectively for Fluency, Flexibility, Originality,
Elaboration and Problem Sensitivity.
Further investigation with the t-test results
provided us that there were 3 cognitive elements
(Fluency, Flexibility and Originality) which are
significantly different between pre-test and post-test,
indicating the enhancements in the abilities of Fluency,
Flexibility and Originality are statistically significant
enough (Table 3). On the other hand, Elaboration (t=-
0.604, p<0.549) and Problem Sensitivity (t=-0.623,

p<0.537) scores are not significantly different between
Fig. 4. Affective modeling with eight emotion elements

5.2 Affective Modeling and its Relation with
Enhanced Design Creativity
In order to measure dynamic characteristics of
students, and to investigate its relationships with the 5
cognitive elements, we incorporated affective
modeling in the creative exercise program.
In the context of computer-assisted learning
context of creative design capabilities, affective
modeling of learners is being done using self-reporting
format. Affective elements composed of joy,
acceptance, apprehension, distraction, sadness,
boredom, annoyance, and anticipation were identified
based on the basic emotion categories proposed by
Plutchik (Plutchik, 2010), which were used in the
affective modeling of the study. The online form of
dialog representing all the affective elements was
devised and presented to students so that the
participants can select one or more affective states
during the experiment. Note that the affection capture
diagram uses identical icons so that other influences
than affective state selection could be isolated in the
interaction of the diagram and the users as the diagram
pops up and prompts affective state selection.
We conducted the online creativity exercise
program with the affective model which is displayed to
students for selections. The affective self-reporting
was done after the learning objectives were given,

after the specific problem statements were given and
after the student problem sessions were done. While
students conduct the exercise program, they are asked
to self-report their affective states using an affective
model diagram as shown in Figure 4.
The collected affective states were used for the
investigation of relationships with the 5 cognitive
elements of design creativity. A machine learning
technique, Association Rules learning was used for
this purpose. Table 4 shows the enhanced design
creativity and its relationships with affective states.
For example, if there is enhanced design creativity
(post test > pre test) then students did not select the
affective states of “Sadness” and “Apprehend”
(Support: 0.66 with Confidence: 0.9). Generally
speaking, the enhanced design creativity is reversely
associated with negative affective states; students did
not select negative affective states when there was
enhancement in design creativity in the post-test.
Rapidminer 5.0 was used in the study for running
Table 4. Association rules between cognitive elements and affective model elements
Premises Conclusion Support Confidence
Distract = false, POST TEST > PRE TEST Apprehend = false 0.62 0.92
Distract = false, POSE TEST > PRE TEST
Sadness = false,
Apprehend = false
0.62 0.92
Sadness = false, Distract = false, POST TEST > PRE TEST Apprehend = fasle 0.62 0.92
POST TEST > PRE TEST Apprehend = false 0.66 0.90
POST TEST > PRE TEST

Sadness = false,
Apprehend = false
0.66 0.90

pre-test and post-test.
Conceptual Design and Cognitive Elements of Creativity: Toward Personalized Learning Supports for Design Creativity 111

machine learning techniques, such as Association
Rules.
6 Conclusion
In the study, we identified the cognitive components of
design creativity and proposed a creativity exercise
program for cognitive elements of design creativity.
This program could be used in helping students
considering their individual needs and contexts, and
enhance design creativity. Five cognitive components
of design creativity were identified, and those are
fluency, flexibility, originality, elaboration and
problem sensitivity. The proposed exercise program
for design creativity was composed of five different
tasks such as making stories, negation, filling black
box, sensitization and diverse classification.
In making stories, the students were required to
produce several different stories by changing order of
three different pictures. The aim of this task was to
improve flexibility, originality and elaboration. The
negation asked students to compulsively negate the
given objects and contrive their alternate purpose or
usage. Accordingly, the students’ flexibility,
originality and problem sensitivity could be enhanced.

In filling black box, the students were supposed to
logically connect given input and output concepts in as
many possible ways within a limited time, and as a
result, the fluency could be improved. The
sensitization asked students to express their feelings on
the given physical objects and abstract concepts
according to five different senses. With this task, the
problem sensitivity could be enhanced primarily and
flexibility secondarily. In diverse classification, the
students were asked to classify the given objects in
several different ways. Therefore, flexibility was
developed and problem sensitivity developed
secondarily.
We conducted an experiment to investigate the
effectiveness of the exercise program for design
creativity cognitive elements. The results show that
there was enhanced creativity, 31 students out of 44
students (70% increases) in terms of the cognitive
elements, after students conducting the proposed
creativity exercise program. Also, the machine
learning results with affective model provided that
there are relations between enhanced creativity in
terms of cognitive elements and negative affective
states, such as Sadness, Apprehend, Distract:. For
example students did not select negative affective
states when there was enhanced creativity.
More rigorous approach is desired to examine what
cognitive elements could be effectively addressed in
each task. This is challenging research because of
uncertain factors and qualitative measurement of data.

However, the research efforts would be helpful for
design creativity education by considering individual's
needs and contexts.
As a future work, thorough investigation of user
data would be helpful in discovering meaningful
results with regard to static and dynamic
characteristics of user. Also, investigation of causal
relationships between enhanced creativity, cognitive
elements and affective states, using machine learning
techniques such as Bayesian learning, will be
important for the identification of factors causing the
enhanced design creativity.
References
de Bono E, (1992) Serious Creativity. Hrper-Collins, London
Goel V, (1995) Sketches of thought. Cambridge, MA: MIT
Press
Guilford JP, Hoepfner R, (1971) The Analysis of
Intelligence. New York: McGraw-Hill
Isaksen SG, Dorval KB, Treffinger DJ, (1998) Toolbox for
Creative Problem Solving. Dubuque, IA: Kendall &
Hunt
Kim MH, Kim YS, Lee HS, Park JA, (2007) An underlying
cognitive aspect of design creativity: Limited
commitment mode control strategy. Design Studies
28(6):585–604
Kim YS, Jin ST, Lee SW, (2010) Relations between design
activities and personal creativity modes, Journal of
Engineering Design, in print
Kim YS, Kim MH, Jin ST, (2005) Cognitive characteristics
and design creativity: An experimental study. In

Proceedings of the ASME International Conference on
Design Theory and Methodology, Long Beach
Kraft U, (2005) Unleashing creativity. Scientific American
Mind 16(1):17–23
Park JA, Kim YS, (2007) Visual Reasoning and Design
Processes. In Proceedings of International Conference
on Engineering Design (ICED), Paris
Plutchik R, (2010) The Nature of Emotions by Plutchik
[Online], />Model/Nature-of-emotions.htm
Treffinger DJ, (1980) Encouraging Creative Learning for the
Gifted and Talented. Ventura, CA: Ventura County
Schools/LTI
Urban KK, (1995) Creativity-A component approach model.
A paper presented at the 11th World Conference on the
Education for the Gifted and Talented, Hong Kong





Analogical Design Computing
DANE: Fostering Creativity in and through Biologically Inspired Design
Swaroop Vattam, Bryan Wiltgen, Michael Helms, Ashok K. Goel and Jeannette Yen
Development of a Catalogue of Physical Laws and Effects Using SAPPhIRE Model
Srinivasan V. and Amaresh Chakrabarti
Measuring Semantic and Emotional Responses to Bio-inspired Design
Jieun Kim, Carole Bouchard, Nadia Bianchi-Berthouze and Améziane Aoussat
Design of Emotional and Creative Motion by Focusing on Rhythmic Features
Kaori Yamada, Toshiharu Taura and Yukari Nagai

DANE: Fostering Creativity in and through Biologically Inspired Design
Swaroop Vattam
1
, Bryan Wiltgen
1
, Michael Helms
1,2
, Ashok K. Goel
1,2
, and Jeannette Yen
2

1
Design & Intelligence Laboratory at Georgia Institute of Technology, USA
2
Center for Biologically Inspired Design at Georgia Institute of Technology, USA
Abstract. In this paper, we present an initial attempt at
systemizing knowledge of biological systems from an
engineering perspective. In particular, we describe an
interactive knowledge-based design environment called
DANE that uses the Structure-Behavior-Function (SBF)
schema for capturing the functioning of biological systems.
We present preliminary results from deploying DANE in an
interdisciplinary class on biologically inspired design,
indicating that designers found the SBF schema useful for
conceptualizing complex systems.
Keywords: Design Creativity, Computational Design,
Biologically Inspired Design, Biomimetic design

1 Introduction
Biologically inspired design uses analogies to
biological systems to derive innovative solutions to
difficult engineering problems (Benyus 1997; Vincent
and Mann 2002). The paradigm attempts to leverage
the billions of biological designs already existing in
nature. Since biological designs often are robust,
efficient, and multifunctional, the paradigm is rapidly
gaining popularity with designers who need to produce
innovative and/or environmentally sustainable designs.
By now there is ample evidence that biologically
inspired design has led to many innovative - novel,
useful, sometimes even unexpected - designs (e.g.,
Bar-Cohen 2006; Bonser and Vincent 2007).
Despite its many successes, the practice of
biologically inspired design is largely ad hoc, with
little systematization of either biological knowledge
from a design perspective or of the design processes of
analogical retrieval of biological knowledge and
transfer to engineering problems. Thus, a challenge in
research on design creativity is how to transform the
promising paradigm of biologically inspired design
into a principled methodology. This is a major
challenge because biology and engineering have very
different perspectives, methods and languages.
We study biologically inspired design from the
perspectives of artificial intelligence and cognitive
science. From our perspective, analogy is a
fundamental process of creativity and models are the
basis of many analogies. Biologically inspired design

is an almost ideal task for exploring and exploiting
theories of modeling and model-based analogies.
We have previously conducted and documented in
situ studies of biologically inspired design (Helms,
Vattam, and Goel 2009). We have also analyzed
extended projects in biologically inspired design
(Vattam, Helms, and Goel 2009). In this paper we
describe the development and deployment of an
interactive knowledge-based design environment
called DANE, which was informed by our earlier
cognitive studies and that is intended to support
biologically inspired design. DANE (for Design by
Analogy to Nature Engine) provides access to a design
case library containing Structure-Behavior-Function
(SBF) models of biological and engineering systems. It
also allows the designer to author SBF models of new
systems and enter them into the library. We present
initial results from deploying DANE in a senior-level
class on biologically inspired design in which teams of
engineers and biologists worked on extended design
projects (Yen et al 2010). The preliminary results
indicate that although we had developed DANE
largely as a design library, in its current state of
development, designers found DANE more useful as a
tool for conceptualizing biological systems.
2 Related Work
Biologically inspired design as a design paradigm has
recently attracted significant attention in research on
design creativity, including conceptual analysis of
biologically inspired design (e.g., Arciszewski and

Cornell 2006; Lenau 2009; Lindermann and Gramann
2004), cognitive studies of biologically inspired design
(e.g., Linsey, Markman and Woods 2008; Mak and
Shu 2008), interactive knowledge-based design tools
for supporting biologically inspired design (e.g.,
Chakrabarti et al. 2005, Sarkar and Chakrabarti 2008;
Chiu and Shu 2007; Nagle et al. 2008), and courses on
biologically inspired design (e.g., Bruck et al. 2007).
116 S. Vattam, B. Wiltgen, M. Helms, A. K. Goel, and J. Yen

Our work on DANE shares three basic features of
similar interactive design tools such as IDEA-
INSPIRE (Chakrabarti et al. 2005, Sarkar and
Chakrabarti 2008). Firstly, both IDEA-INSPIRE and
DANE provide access to qualitative models of
biological and engineering systems. Secondly, both
IDEA-INSPIRE and DANE index and access the
models of biological and engineering systems by their
functions. Thirdly, both IDEA-INSPIRE and DANE
use multimedia to present a model to the user
including structured schema, text, photographs,
diagrams, graphs, etc.
However, our work on DANE differs from IDEA-
INSPIRE and similar tools in three fundamental
characteristics. Firstly, the design and development of
DANE is based on our analysis of in situ cognitive
studies of biologically inspired design (Helms, Vattam,
and Goel 2009; Vattam, Helms and Goel 2009).
Secondly, insofar as we know, IDEA-INSPIRE has
been tested only with focus groups in laboratory

settings. In contrast, we have introduced DANE into a
biologically inspired design classroom. This is
important because from Dunbar (2001) we know that
the analogy-making behavior of humans in naturalistic
and laboratory settings is quite different: in general,
humans make more, and more interesting, analogies in
their natural environments. Thirdly, while IDEA-
INSPIRE uses SAPPhIRE functional models of
biological and engineering systems, DANE uses
Structure-Behavior-Function (SBF) modeling (Goel,
Rugaber and Vattam 2009). This is important because
SBF models were developed in AI research on design
to support automated analogical design (e.g., Bhatta
and Goel 1996, Goel and Bhatta 2004). Thus, in the
long term it should be possible to add automated
inferences to DANE.
An SBF model of a complex system (1) specifies
the structure, functions, and behaviors (i.e., the causal
processes that result in the functions) of the system, (2)
uses functions as indices to organize knowledge of
behaviors and structures, (3) represents behavior as a
series of states and state transitions that are annotated
with causal explanations, (4) organizes the knowledge
in F  B  F  B …  F(S) hierarchy, and (5)
provides an ontology for representing structures,
functions and behaviors. Other researchers have
developed similar functional models e.g., Kitamura et
al. 2004 and Umeda et al. 1996.
3 The Design By Analogy to Nature Engine
In the long term, DANE is intended to semi-automate

analogical retrieval and transfer in biologically
inspired design. Presently, DANE interactively
facilitates biologically inspired design by (1) helping
designers find biological systems that might be
relevant to a given engineering design problem, (2)
aiding designers in understanding the functioning of
biological systems so that they can extract, abstract
and transfer the appropriate biological design
principles to engineering design problems, and (3)
enabling designers to construct and refine SBF models
of biological and engineering systems.
DANE employs a client-server architecture with a
centralized design repository on the server-side. Each
client is a thin client whereby all data is stored,
updated, and recalled from the server. This
architecture supports simultaneous access by multiple
users and allows users to browse or edit the most
current version of the repository.
DANE is a distributed Java application running on
the Glassfish application server. Data is stored in a
MySQL database, and we use EJB technology to
handle persistence and connection pooling. Users
access the application by going to a launch website
that utilizes Java Web Start to both download and
execute the application as well as apply any updates
that have been made since the user last launched the
application.
DANE’s library of SBF models of biological and
engineering systems is growing. In early fall of 2009,
when we introduced the system into a biologically

inspired design classroom, the library contained about
forty (40) SBF models, including twenty two (22)
“complete” models of biological systems and
subsystems. The remaining were either SBF models of
engineering systems or only partial models of
biological systems. Biological systems in DANE were
at several levels of scale from the sub-cellular to organ
function to organism.
Systems are indexed by system-function pairs and
retrieved by function name (e.g., “flamingo filter-feeds
self”), by subject (e.g., “flamingo”), and/or by verb
(e.g., “filter-feeds”). Function names often include
additional specificity with regard to the objects upon
which the function acts. In this case the flamingo is
feeding itself. Upon selecting a system-function pair,
users are presented with a multi-modal representation
of the paired system-function (e.g. the “flamingo filter-
feeds self” SBF model). For example, in DANE a
system can be represented in text descriptions and
images, as well as through visualizations of behavior
and structure models. Example text and image
modalities for the “flamingo filter-feeds self” model
can be seen in Figure 1.
Briefly, this model describes how a flamingo uses
its tongue to create negative pressure in its slightly
open mouth to draw water in, closes its mouth, and
then uses its tongue to force the water out through a
filter-system composed of comb-like lamellae and
DANE: Fostering Creativity in and through Biologically Inspired Design 117

mesh. The lamellae trap the food, which is then drawn
into the flamingo’s esophagus in the next cycle.
Behavior and structure parts of the SBF models are
themselves represented as directed graphs, which may
be annotated with text descriptions and images. The
nodes and edges represent either structural elements
and connections (for structure models) or states and
transitions (for behavior models), respectively. We
provide an example of a partial behavior model, this
time for the system “kidney filters blood,” in Figure 2.
Note that the annotations on the transitions in this
figure are labeled with short-hand that denotes their
type: [FN] X identifies that a transition occurs because
of some sub-function X, and [STR_CON] X Y
identifies that a transition occurs because of the
connection between some structural component X and
another structural component Y.
This “kidney filters blood” partial behavior model
(a component of the larger SBF model) describes the
movement of blood through the kidney through
smaller and smaller vessels until the blood arrives at
the nephron, where the filtration process takes place.
Although in DANE the complete behavior model
would be displayed, due to space constraints we only
show in our figure a few states and transitions in this
behavior. The sub-function “nephron purifies blood”
serves as an index to yet another SBF model that
describes this complex lower-level process in more
detail. This provides an example of how SBF models
are nested through function.

Additionally, each system is visually connected to
other systems with which it shares a sub or super-
function relationship. This functional hierarchy is
represented as an interactive graph with nodes
representing systems and edges representing the
sub/super relationships. Users may navigate between
systems by double-clicking on a node. Figure 3
illustrates the functional hierarchy graph for the
system “sliding filament model” and shows the
browsing window with a few systems displayed,
including the flamingo filter-feeding self function. The
“sliding filament model” describes how muscle fibers
contract, and thus the model is connected to a number
of higher level animal functions (e.g. “flamingo filter-
feeds self” and “basilisk lizard walks on water”), and
is connected to a number of lower level molecular
functions related to myosin and ATP. We can see in
this one example how SBF models operate and
connect functions at many scales.
By presenting complex systems in the SBF
schema, which places an emphasis on the causal
relationships within each system, and by making
explicit the function/sub-function relationships
between systems, we hypothesize that biologists and
engineers will understand the systems in a way that (a)
helps them identify systems that are relevant to their
design problem and (b) is transferable to a design
solution. For example, an engineer might scan models
in DANE until he/she comes across a system that has a
similar initial and objective state (a function) that

matches his/her design problem. Then, by inspecting
the structure and behavior of that system, the engineer
might formulate a technological solution that
implements a similar set of behaviors.
While SBF models can represent systems across
multiple levels of scale and abstraction, and across the
two domains of biology and engineering, the issue of
knowledge engineering remains problematic. In

Fi
g
. 1. Exam
p
le of a
m
ulti-modal
m
odel of a flamin
g
o’s filte
r
-feedin
g
a
pp
aratus in DANE
118 S. Vattam, B. Wiltgen, M. Helms, A. K. Goel, and J. Yen

particular, we found that constructing a “complete”
SBF model of a complex biological system requires

between forty (40) and one hundred (100) hours of
work. The process of understanding the biological
system (e.g. the kidney), modeling it in the SBF
language, discovering faults in the model or in the
modeler’s understanding, and iterating over this
process consumed a large majority of the time. We
estimate that just entering a complete model into
DANE required somewhat less than 25% of the overall
time cost.
4 Application Context
We deployed DANE in the Fall 2009 semester session
of ME/ISyE/MSE/PTFe/BIOL 4803, a project-based,
senior-level, undergraduate course taught by biology
and engineering faculty affiliated with Georgia Tech’s
Center for Biologically Inspired Design (Yen et al.
2010). The class composition too was
interdisciplinary, comprising of 15 biology students,
11 mechanical engineering students, and 14 students
from a variety of academic disciplines including
biomedical engineering, chemical engineering,
industrial engineering, material science, mathematics,
and a few other engineering fields.
The course has three components: lectures, found
object exercises, and a semester-long biologically
inspired design team project. In the design project,
teams of 4-6 students were formed so that each team
would have at least one biology student and students
from different schools of engineering. Each team was
given a broad problem in the domain of dynamic,
adaptable, sustainable housing such as heating or

energy use. Teams are expected to refine the problem
and then design a biologically inspired solution based
on one or more biological sources to solve it. All
teams presented their final designs during the end of
the class and submitted a final design report.
The class is taught without any aids for design or
research. Students are encouraged to perform their
own research on biological systems through resources

Fig. 3. List of functions and a functional hierarchy for
“Sliding Filament Model” in DANE.

Fig. 2. Partial behavior model of “Kidney filters blood“ in DANE